Research has shown that reducing SUDs through effective treatment leads to a reduction in criminal activity\(^{[1]}\). However, most evidence comes from developed countries, and results from the Latin American context are largely unknown\(^{[2]}\). The structural, economic, epidemiological context and substance use treatment (SUT) policy response are different in this region, making the question about SUT effectiveness through locally based data relevant\(^{[3]}\).
We analyze Chile as a case study and examine the impact of SUT on the prevention of contact with the criminal justice system (CJS) in the short (3 and 6 months), middle (1 year), and long term (3 years). Hypothesis: Patients who complete treatment have lower probabilities of being in contact with CJS compared to patients who do not complete, although this effect may decrease as observation time passes.
This research relies on a population-based record-linkage retrospective cohort design. We used a deterministic linkage process (using encryption of the Chilean Unique National ID) to merge electronic records of individuals in publicly funded Chilean SUT programs with the Prosecutor’s Office data at the national level between 2010 and 2019. This research is approved by the Griffith University Human Research Ethics Committee (GUHREC) (GU Ref No: 2022/919).
We described the cumulative incidence rate and incidence rate ratio (IRR) of contact with the CJS (offenses that ended with a condemnatory sentence and of offenses that ended with imprisonment after baseline treatment outcome), and its variation by baseline treatment outcome: Treatment completion, Late (>= 3months) & Early Discharge (within the first 3 months of treatment).
We calculated the association between Baseline treatment outcome and Contact with CJS through Royston-Parmar models while adjusting for several covariates and obtained standardized survival curves and restricted mean survival times (RMST) through the stpm2 command in Stata\(^{[4]}\). Missing data was imputed using multiple imputation with regression trees from missRanger R package\(^{[5]}\). Secondary analyses included e-values of the strength of confounding needed to take away the associations between treatment outcome and contact with CJS. Codes are available at bit.ly/40cMATs. Covariates are listed below:
| Time | Complete Tr. | Late Disch. | Early Disch. | Comp. vs Late | Comp. vs Early | Early vs Late |
|---|---|---|---|---|---|---|
| Probs. | ||||||
| 3_mths | 96.9 (96.7,97.1) | 94.7 (94.5,94.9) | 93.8 (93.5,94.1) | -2.2 (-2.5,-1.9) | -3.1 (-3.4,-2.7) | .9 (.5,1.2) |
| 6_mths | 94.3 (94,94.6) | 90.6 (90.4,90.9) | 89.3 (88.9,89.8) | -3.6 (-4,-3.3) | -4.9 (-5.4,-4.4) | 1.3 (.8,1.8) |
| 1_yr | 90 (89.6,90.4) | 84.4 (84.1,84.8) | 82.8 (82.3,83.3) | -5.5 (-6,-5) | -7.2 (-7.9,-6.5) | 1.7 (1,2.3) |
| 3_yrs | 79.4 (78.8,80) | 70.9 (70.4,71.3) | 69 (68.3,69.7) | -8.6 (-9.3,-7.9) | -10.4 (-11.4,-9.5) | 1.9 (1,2.7) |
| 5_yrs | 73.4 (72.7,74.2) | 63.9 (63.4,64.5) | 62.2 (61.4,63.1) | -9.5 (-10.3,-8.7) | -11.2 (-12.3,-10.1) | 1.7 (.8,2.7) |
| RMST | ||||||
| 3_mths | .251 (.25,.251) | .247 (.247,.248) | .246 (.246,.247) | -.002 (-.002,-.002) | -.003 (-.003,-.002) | .001 (.001,.001) |
| 6_mths | .494 (.493,.495) | .483 (.482,.484) | .479 (.478,.48) | -.011 (-.012,-.01) | -.015 (-.017,-.013) | .004 (.003,.006) |
| 1_yr | .962 (.959,.965) | .928 (.925,.93) | .916 (.912,.92) | -.034 (-.038,-.031) | -.046 (-.051,-.041) | .012 (.007,.016) |
| 3_yrs | 2.622 (2.61,2.635) | 2.442 (2.433,2.452) | 2.394 (2.379,2.41) | -.18 (-.195,-.164) | -.228 (-.248,-.207) | .048 (.029,.067) |
| 5_yrs | 4.172 (4.148,4.197) | 3.807 (3.788,3.825) | 3.722 (3.692,3.752) | -.366 (-.395,-.336) | -.45 (-.491,-.41) | .085 (.049,.121) |
| Time | Complete Tr. | Late Disch. | Early Disch. | Comp. vs Late | Comp. vs Early | Early vs Late |
|---|---|---|---|---|---|---|
| Probs. | ||||||
| 3_mths | 99.5 (99.5,99.6) | 99.1 (99,99.2) | 98.8 (98.7,98.9) | -.5 (-.6,-.4) | -.7 (-.9,-.6) | .3 (.1,.4) |
| 6_mths | 99.1 (99,99.3) | 98.4 (98.3,98.5) | 98 (97.8,98.1) | -.8 (-.9,-.6) | -1.2 (-1.4,-1) | .4 (.2,.6) |
| 1_yr | 98.4 (98.3,98.6) | 97.2 (97.1,97.4) | 96.6 (96.3,96.8) | -1.2 (-1.4,-1) | -1.9 (-2.2,-1.5) | .6 (.4,.9) |
| 3_yrs | 96.4 (96.1,96.7) | 94.3 (94,94.5) | 93.2 (92.8,93.6) | -2.2 (-2.5,-1.8) | -3.2 (-3.7,-2.8) | 1.1 (.6,1.5) |
| 5_yrs | 94.9 (94.5,95.3) | 92.3 (92,92.6) | 91 (90.5,91.4) | -2.6 (-3.1,-2.2) | -4 (-4.6,-3.3) | 1.3 (.8,1.9) |
| RMST | ||||||
| 3_mths | .254 (.254,.254) | .253 (.253,.253) | .253 (.253,.253) | -.001 (-.001,0) | -.001 (-.001,-.001) | 0 (0,.001) |
| 6_mths | .507 (.506,.507) | .505 (.504,.505) | .503 (.503,.504) | -.002 (-.003,-.002) | -.003 (-.004,-.003) | .001 (.001,.002) |
| 1_yr | 1.01 (1.008,1.011) | 1.002 (1.001,1.003) | .998 (.997,1) | -.007 (-.009,-.006) | -.011 (-.013,-.009) | .004 (.002,.006) |
| 3_yrs | 2.931 (2.925,2.936) | 2.889 (2.884,2.893) | 2.867 (2.86,2.875) | -.042 (-.049,-.035) | -.063 (-.073,-.054) | .022 (.013,.031) |
| 5_yrs | 4.878 (4.865,4.89) | 4.786 (4.777,4.796) | 4.74 (4.724,4.755) | -.091 (-.107,-.076) | -.138 (-.158,-.117) | .046 (.027,.065) |
[1] M. Prendergast, D. Podus, E. Chang, et al. “Erratum to The effectiveness of drug abuse treatment: a meta-analysis of comparison group studies”. In: Drug and Alcohol Dependence - DRUG ALCOHOL DEPENDENCE 84 (sept.. 2006), pp. 133-133. DOI: 10.1016/j.drugalcdep.2006.02.002.
[2] H. Klingemann. “Successes and Failures in Treatment of Substance Abuse: Treatment System Perspectives and Lessons from the European Continent”. In: Nordisk Alkohol- and Narkotikatidskrift 37.4 (2020), pp. 323-37.
[3] M. Mateo Pinones, A. González-Santa Cruz, R. Portilla Huidobro, et al. “Evidence-based policymaking: Lessons from the Chilean Substance Use Treatment Policy”. En. In: Int. J. Drug Policy 109.103860 (nov.. 2022), p. 103860.
[4] P. Lambert. STPM2: Stata module to estimate flexible parametric survival models. Statistical Software Components, Boston College Department of Economics. feb.. 2010. URL: https://ideas.repec.org/c/boc/bocode/s457128.html.
[5] M. Mayer. “missRanger: Fast Imputation of Missing Values”. (2023). R package version 2.2.0. URL: https://github.com/mayer79/missRanger.